Thursday, May 19, 2016

Impulse and PvP

Impulse and PvP

Tapion, a very read-dependent Falcon player of lore, gave me two tendencies to note in the PvP. I have added two additional rules of thumb for a list of four.

1) What a player does first he will do most often.
2) If you directly punish something, they will not do that again in the next instance (if the event passes unrecognized then ignore this rule).
3) Players will use ~3 options at most in any given situation.
4) Decisions are prompted by position and/or relative frame advantage.

While this methodology has it’s most direct application in tech-chasing, high-level players actively use these guidelines to predict DI, double jumps, exit routes, etc.

Let’s look at these tendencies individually.

1) What a player does first he will do most often.
This is pretty straightforward. With no information to go off of, a decision is made impulsively or because it is comfortable. This means that whatever action is taken is impulsive or comfortable. In a game as cut-throat as Melee no one opens with intentional risk.

2) If you directly punish something, they will not do that again in the next instance (if the event passes unrecognized then ignore this rule).
Unless a player is oblivious to the interaction, they attempt to learn from and control their previous decision. For whatever reason (I’m not up on the psychology enough to say exactly) this means that they choose something else the next time they recognize the decision. However, if they do not recognize the decision as a choice then Rule #1 stands. Humans are profoundly terrible at randomizing their behavior. Every decision has a source and you probably just located it.

3) Players will use ~3 options at most in any given situation.
Normally your character will have one or two options that are clearly, mechanically better than then alternatives. These decisions are reinforced every time you make that decision. As it plays out over and over it becomes a habit. I noticed this while I was researching Puff on Shield. No player uses every option from shield. Even in the cases where they attempt to mix things up they are just going to commit to 1 of about 3 that they've internalized as "sufficiently good." This is a general trend that goes beyond techs and OoS options and into various positions etc and allows for massive adaptation by a mindful player. Obviously you can net a clean stock off of a clean read (especially as Jigglypuff) but often times knowing what an opponent won’t do is exceedingly useful. There are only so many other options that they can or will commit to. A player afraid to tech roll in will stay in the vicinity of tech in place/out, a much smaller and normally reactable range. A Fox that won’t FH nair enables a number of Puff’s otherwise overly risky nair/fair patterns that can consistently beat out many of his other options.

4) Decisions are prompted by position and/or relative frame advantage.
People will react differently to you standing in front of your shield vs behind it. They techroll differently depending on where they hit relative to center stage. Generally, impulsive decisions are all directly related to the perceived threat more than the moves etc involved.
For frame advantage refer to

Obviously this is not an exact science. It only brushes against conditioning or confusing non-zero-sum theory that I don't intend to read (lol). However after careful thought and testing I believe that these guidelines are simple and comprehensive enough to have massive net gains over time that can supplement a strong gameplan.

If you can react then don't bother with this shit and take your guaranteed stuff


  1. If this analysis is true, which I believe it largely is, it raises the question of how to use this information. In terms of predicting your opponents' tendencies, the application is pretty straightforward, but for improving your own habits, it's not quite clear. These tendencies may not have been thought out logically, but that doesn't necessarily mean they are bad frameworks to play within.

    The first two seem heavily situational. As with classic Yomi/RPS examples, doing certain options is good at certain levels of play. Teching in place a second time after getting punished for doing it once may suck vs. low level opponents who repeat punishes until their opponent demonstrates a willingness to adapt, but at high level, players often predict mix ups after punishing something because they give their opponent the benefit of the doubt due to their skill. In that case, teching in place a second time would clearly be the best decision. These impulses are messy because of how meta-dependent they are.

    The third and fourth are more complex. Do players tend to limit themselves to 3 options because of human limitations and natural tendencies? So much of fighting games tend to be simplified to RPS situations even when there are more than 3 options, so I wouldn't be surprised if this was a human element, but it seems equally likely that 3 just tends to be the appropriate number of options you should choose from. Adding a 4th option to your repertoire may make you less predictable, but at what cost? You are now subjecting yourself to the usage of worse options which may lead to significantly harder punishes when they occasionally do predict you. Are we better off sticking with 3 really good options that limit risk and maximize reward, or is this our evolutionary monkey brains failing to problem solve an unnatural interaction? If our opponent is USED to playing opponents with 3 main options, does adding a 4th option throw a wrench into their own strategy more than our own?

    There's also the issue of having to learn the depth of another option. It's hard enough trying to master mixing up the ins and outs of 3 options. Mastering 4 or 5, each with an entire branch of sub-options extending from it, is no small task. For every option you add in Melee, we could be talking about hundreds of extra hours you would need to dedicate specifically to those options in order to really understand and apply them efficiently. Imagine if Melee were patched with some sort of quick tech in place option that was half the length of a normal tech, but with no invul. This simple addition would have a drastic ripple effect through the rest of the game, and usage of the quick-tech would no doubt evolve over time just as much as the other tech options have.

    (I realize I'm asking a lot of hypotheticals that are nearly impossible to answer, but I figured I'd think out loud since this post got me brainstorming.)

    1. (^_^*)
      Yeah it gets pretty weird pretty quickly in terms of yomi. Especially because you're rarely sure just how familiar your opponent is with the yomi you're trying to play.
      My goal with this and a lot of what I put up is to create a cleaner but still comprehensive framework/vocabulary so that thinking about the next steps is easier and more effective.